ggml-cpu: optimize ggml_vec_dot_bf16 for s390x

[taronaeo]

Similar to #18837, this pull request integrates the SIMD instruction set for BF16 on the s390x platform. We notice a 154.86% performance improvement for Prompt Processing. No performance difference was noticed for Token Generation.

Before SIMD Benchmark

model	size	params	backend	threads	mmap	test	t/s
granite 3B BF16	4.72 GiB	2.53 B	CPU	1	1	pp512	0.29 ± 0.00
granite 3B BF16	4.72 GiB	2.53 B	CPU	1	1	tg128	0.03 ± 0.00

build: 423cf0b (8029)

After SIMD Benchmark

model	size	params	backend	threads	mmap	test	t/s
granite 3B BF16	4.72 GiB	2.53 B	CPU	1	1	pp512	2.28 ± 0.20
granite 3B BF16	4.72 GiB	2.53 B	CPU	1	1	tg128	0.03 ± 0.00

build: b4bc24e (8004)

Verification

This PR was tested against the IBM Granite 3.3 2B Instruct BF16 model.

test-quantize-fns

$ build/bin/test-quantize-fns

Testing f32
Testing f16
Testing q4_0
Testing q4_1
Testing q5_0
Testing q5_1
Testing q8_0
Testing q8_1
Testing q2_K
Testing q3_K
Testing q4_K
Testing q5_K
Testing q6_K
Testing q8_K
Testing iq2_xxs
Testing iq2_xs
Testing iq3_xxs
Testing iq1_s
Testing iq4_nl
Testing iq3_s
Testing iq2_s
Testing iq4_xs
Testing i8
Testing i16
Testing i32
Testing i64
Testing f64
Testing iq1_m
Testing bf16
Testing tq1_0
Testing tq2_0
Testing mxfp4

[taronaeo]

@CISC I'm not sure who to ping to review this... Any chance you could review it? :)

[CISC]

I can't verify this, but it looks good in general and CI passes.